AI 中文总结
本研究通过对VLA策略的闭环审计，发现任务向量减法作为快速模型编辑方法存在局部性脆弱问题，会引发目标控制分离、抗性或全局崩溃等多种不良情况，凸显了闭环评估的必要性。
AI 中文摘要
任务向量算法提供了一种修改模型的闭式方法，但其在闭环机器人控制中的行为局部性仍不明确。我们对多任务视觉-语言-动作（VLA）策略中按技能进行的任务向量减法开展了目标与控制审计。在全部10项LIBERO-Goal技能中，减法产生了三类性质不同的情况：5项技能出现目标-控制分离，3项表现出抗性，2项发生全局崩溃。在保留的初始状态下，5项可抑制目标的成功率仍为0%；然而，基线归一化的平均控制保留率仅为52%，且每项抑制目标的编辑都会实质性损害至少一项名义上不相关的控制。额外的Goal面板显示，在测试的采用连续回归、离散标记和流匹配动作头的策略中存在分离，而在测试的Spatial面板上未观察到清晰分离，在测试的Object和Long-horizon面板上则出现控制崩溃。平均任务向量余弦无法解释这种变化。匹配范数控制识别出一个Goal锚点周围的局部符号不对称，而多向量结果随锚点和规模变化。保留感知梯度基线提供了依赖数据的比较指标，但需要移除时的数据和优化；减法仅在编辑时无数据和梯度依赖，前提是有预先计算的专家增量。最后，单技能重训练探测结果与行为掩蔽一致，而非经认证的遗忘。这些结果将任务向量减法描述为一种快速但脆弱的干预措施，并强调在评估具身模型编辑的局部性时，需开展闭环目标与控制评估。
英文摘要
Task-vector arithmetic offers a closed-form way to modify a model, yet its behavioral locality remains unclear in closed-loop robot control. We present a target-and-control audit of per-skill task-vector subtraction from multitask vision-language-action (VLA) policies. Across all ten LIBERO-Goal skills, subtraction produces three qualitatively different regimes: target-control separation for five skills, resistance for three, and global collapse for two. On held-out initial states, the five suppressible targets remain at 0% success; however, mean baseline-normalized control retention is only 52%, and each target-suppressing edit materially harms at least one nominally unrelated control. Additional Goal panels show separation across tested policies with continuous-regression, discrete-token, and flow-matching action heads, whereas we observe no clean separation on Spatial and control collapse on the tested Object and Long-horizon panels. Mean task-vector cosine does not account for this variation. A matched-norm control identifies a local sign asymmetry around one Goal anchor, while multi-vector outcomes vary with anchor and scale. Retain-aware gradient baselines provide data-dependent comparators but require removal-time data and optimization; subtraction is data- and gradient-free only at edit time, assuming precomputed expert deltas. Finally, a single-skill relearning probe is consistent with behavioral masking, not certified unlearning. These results characterize task-vector subtraction as a fast but brittle intervention and underscore the need for closed-loop target-and-control evaluation when assessing locality in embodied model editing.
Comments28 pages, 14 figures, 40 tables. Preprint
